Plants. A total of 10 cultivars of sorghum – eight hybrids and two inbred lines - were used for this study. Eight of these were from the Texas A&M AgriLife Research Sorghum Improvement Program. Four of these cultivars (2 hybrids, ATx3408/RTx436 and ATx3409/RTx436, and two inbred lines, BTx3408 and BTx3409) exhibit direct resistance against the greenbug and sorghum aphids, whereas four other hybrids (ATx631/RTx436, ATx645/RTx436, BTx631, and BTx645) are susceptible to aphid infestations (Mbulwe et al. 2016, Armstrong et al. 2015). The remaining two commercial hybrids are products of Bayer®: DKS38-16 and DKS37-07 which are respectively susceptible and resistant to aphid infestations (Szczepaniec 2018). Experimental and commercial sorghum lines were selected based on differences in their lineage, such that our experiments would capture some genetic diversity in indirect defense traits.
Two seeds were planted per pot (7 cm in diameter x 6.5 cm depth) in Pro-Mix LP15 growing medium (Premier Tech Horticulture, Quakertown, Pennsylvania) and thinned to one plant after about 1 week. Plants were grown at 28–32℃ with supplemental light (16:8 light:dark cycle). Sorghum seedlings used in the volatile collections and olfactometer bioassays were grown until there were 4–6 true leaves, taking approximately 3 weeks after planting. All sorghum was watered ad libitum when the soil was dry.
Insects. Sorghum aphid colonies were established with aphids collected from sorghum in College Station, Texas in June of 2020. Colonies were maintained on DKS38-16 sorghum cultivar plants in a growth chamber (Percival Scientific Inc., Perry, Iowa) at 30℃, 40% RH, and 12:12 light:dark photoperiod. New sorghum plants were added to the colony every other day to maintain aphid abundances throughout the study. For the choice assays, we selected natural enemies that are highly correlated with sorghum aphids in sorghum (Maxson et al. 2019, Szczepaniec 2018, Faris et al. 2022). Typically, adult lacewings are used to assess natural enemy attraction to HIPVs. In this study, we chose to use larval lacewings for several reasons: 1) Chrysoperla rufilabris larvae, not adults, are predators of aphids, 2) larval lacewings can be bought in large, cost-effective numbers that are easy to maintain, and 3) our previous research has demonstrated that larval C. rufilabris are capable of significant suppression of aphids (Hewlett et al. 2019).
A. nigritus mummies were collected from sorghum aphid colonies on sorghum in College Station, TX in June of 2020. The parasitoid colony, maintained on sorghum aphid-infested DKS38-16 plants, was kept in an environmental chamber 16:8 light:dark photoperiod, 23–26℃, and 70–75% RH. Mummies were placed in petri dishes and checked daily for emergence. C. rufilabris eggs were obtained from ARBICO Organics (Oro Valley, AZ, USA). The lacewings were reared in a growth chamber set to 27, 75% RH, and 12:12 light:dark photoperiod. Eggs were separated into plastic cups and upon hatching, larvae were separated using a paintbrush. Larvae were then individually placed into their own plastic cups and fed approximately 100 aphids per day. Larvae were reared to the third instar before use in the olfactometer assays.
Olfactometer Assays. To assess natural enemy preference, a Y-tube olfactometer (ARS Gainesville, FL, OLFM-ADS-2AFM1C) was used. Two natural enemy species, parasitoid A. nigritus and predator C. rufilabris were used for this study. After assessing parasitoid responses to all ten of the sorghum cultivars, we selected two cultivars of interest (DKS37-07 and ATx3408/RTx436) to assess attraction of a predator and determine if this would match that of the parasitoid. Each replicate of the choice assay featured a pair of sorghum plants of the same cultivar, one with aphids and one without aphids. Sorghum plants at the 4–6 leaf stage were infested with 100 aphids 48 h before the start of the choice experiments. Aphids were re-counted 24 h after initial infestation and aphids were supplemented or removed to ensure that a consistent number remained on the sorghum plant. Virgin female adult A. nigritus (16 − 14 h old) and third instar C. rufilabris were starved 24 h before the start of the experiment. Airflow entering the olfactometer was drawn through activated charcoal, purified before entering the y-tube, and maintained at 15 psi. A pair of potted plants of the same cultivar with and without aphids were placed in the two 4 L glass chambers. An individual natural enemy was then placed individually into the base of the y-tube. The natural enemy was observed for 5 min and the arm choice was recorded. If the individual did not select a path (a path was considered selected once the natural enemy is ¾ way up an arm) after 5 min, “no choice” was recorded. After each trial, the y-tube was rinsed with 70% ethanol, rinsed with distilled water, and blown dry with a hair dryer. Placement of infested and un-infested plants were alternated between the two glass jars.
Volatile Collections. All 10 of the sorghum cultivars were used for volatile emission collections. Sorghum seedlings were infested with 100 aphids or remained aphid-free. To ensure the correct number of aphids were present, plants assigned to the herbivory treatment were kept in a single rearing tent and aphids were re-counted 24 h after infestation. If needed, aphids were added or removed so that all herbivory treatment plants had approximately 100 aphids. Rearing tents were kept at least one meter apart to prevent volatile-mediated interactions among plants. Collections were performed in walk-in growth rooms (28°C; 16:8 light:dark cycle; 25% RH). Individual plants were placed in heat-resistant oven bags (Reynolds Consumer Products, Lake Forest, Illinois) that were sealed at the base of the plant with a twist tie. Prior to this, oven bags were baked (150℃, 8 h) to remove caprolactam volatiles (Stewart-Jones and Poppy 2006, Derstine et al. 2020). Air was filtered and pumped into the oven bags at 200 mL min− 1 while a vacuum pump pulled air out of the oven bags at the same time at 200 mL min− 1. Volatiles were pulled into volatile filter traps containing 45 mg HayeSep® Q (Hutchison Hayes Separation Inc., Houston, Texas) and volatile collections lasted 8 h (09:00–17:00). Volatiles were eluted from the volatile filter traps with 150 µl of dichloromethane and each sample contained 5 µl (400 ng) of a nonyl acetate standard. Plants were cut at the base of the stem and wrapped in aluminum foil to be dried in an oven (150℃, 24 h) and dry mass was recorded for each plant.
Volatiles were quantified using an Agilent 7890 gas chromatograph (GC) and 5977B mass spectrometer. A 1 µl aliquot was injected into the column (HP-5MS 30 m x 0.250 mm-ID, 0.25 µm film thickness, Agilent Technologies). After injection, the column was held at 40℃ for 5 min, then increased 20℃/min− 1 to 250℃. Helium was used as the carrier gas. The mass spectrometer used electron impact ionization to ionize compounds at 70 eV and produced mass spectra by scanning from 40 to 300 m/z. Tentative identification of volatile compounds were made by comparison with mass spectral libraries (NIST17, Adams2) and confirmed where possible using authentic standards (Grunseich et al. 2020). The relative abundance of each compound in ng was calculated based on the internal standard. Concentrations of volatile compounds were determined by dividing this number by the dry weight of the plant.
Statistical Analysis. All analyses were completed using R (R Version 4.2.2, R Core Team 2022). To test the null hypothesis that natural enemies had no preference for either aphid-infested or un-infested plants (i.e., a preference for one of the two olfactometer arms), chi-square tests were conducted. Total volatile emissions were calculated for each sample and log transformed to meet assumptions of homoscedasticity and normality. A two-way ANOVA was used to determine the effects of cultivar and aphid infestation on total volatile emission. To visualize and qualitatively compare HIPV blends, non-metric multidimensional scaling (NMDS) and Bray-Curtis dissimilarity metrics with the R package vegan were used (Oksanen et al. 2020). A permutational analysis of variance (PERMANOVA; number of permutations performed = 999) was used to test the hypothesis that volatile blends differed among aphid-infested sorghum cultivars using the R package vegan and the function adonis (Oksanen et al. 2020, Grunseich et al. 2020).
Random forest analyses and variable selection using the Boruta algorithm (packages: randomForest and Boruta) were used to determine which individual compounds differed most between treatments (infested and un-infested plants of the same cultivar) and among all aphid-infested cultivars (Liaw and Winer 2002, Kursa and Rudnicki 2010). Random forest analysis is a multivariate technique that creates many decision trees via bootstrapping (Breiman 2001) and is used for the classification of volatile datasets and identification of volatiles that best discriminate among groups or treatments (Ranganathan and Borges 2010). Random forest is useful for volatile datasets, as it can utilize non-parametric data, is robust to correlations among variables, can be used when there are more variables than number of samples, and is not biased toward abundant compounds (Ranganathan and Borges 2010, McCormick et al. 2014). Volatile compounds that were predicted to be different among treatments (infested vs. un-infested plants) were then analyzed by Mann-Whitney tests to determine if the amounts of volatiles changed after induction with aphid herbivory. Volatiles predicted to be different among aphid-infested cultivars were further analyzed by Kruskal-Wallis and mean separation Dunn tests to determine if the amounts of volatiles different among cultivars.